Add scattering loss estimations to ModeSolver #2892
Description
Feature Request: Modesolver-based scattering loss from rough surfaces
Radiated loss from rough surfaces on a waveguide is a "traditionally" difficult thing to calculate accurately. The common Payne and Lacey model really doesn't work well for 2D structured waveguides with high-index contrast, often not even following correct scaling trends. Brute-force calculation through FDTD, even with a high-speed solver such as FlexCompute's, is impractical: the mesh resolution required to resolve ~nm scale RMS roughness is prohibitive, and the length scales over which we care about losses (typically ~cm) are insanely expensive to simulate outright. (there are some more realistic FDTD-based approaches we can discuss but what I'm requesting I think is preferable from a user experience standpoint)
It would be fantastic if Tidy3D's modesolver implemented one of the more rigorous methods for calculation of this radiative loss based on RMS roughness + correlation length of roughness. Two examples come to mind:
Tymon Barwicz's approach and Chris G. Poulton's approach.
I also recently saw this paper from Graz which may or may not be suitable.
The first is implemented in eModePhotonix.
Since Tymon seems to have a strong link to FlexCompute from the GoogleX/GDSfactory days, maybe he'd be willing to help out here? :) I have his old Matlab codes, but being from 2005 and with a bunch of custom-written stuff I could never get it working. I imagine implementing one of these two approaches should be much more straightforward given modern tools in Python.